Re: [PATCH v3 2/3] spi: Add a mechanism to use the fwnode name for the SPI device

[Andy Shevchenko <andy.shevchenko@xxxxxxxxx>]

On Mon, Apr 8, 2024 at 4:52 PM Charles Keepax
<ckeepax@xxxxxxxxxxxxxxxxxxxxx> wrote:
> On Wed, Apr 03, 2024 at 11:29:27PM +0300, Andy Shevchenko wrote:
> > Fri, Mar 29, 2024 at 11:47:29AM +0000, Charles Keepax kirjoitti:

...

> > >     struct acpi_device *adev = ACPI_COMPANION(&spi->dev);
> > > +   struct fwnode_handle *fwnode = dev_fwnode(&spi->dev);
> > > +
> > > +   if (spi->use_fwnode_name && fwnode) {
> > > +           dev_set_name(&spi->dev, "spi-%s", fwnode_get_name(fwnode));
> > > +           return;
> > > +   }
> > >
> > >     if (adev) {
> > >             dev_set_name(&spi->dev, "spi-%s", acpi_dev_name(adev));
> >
> > This should be something like this
> >
> >       struct device *dev = &spi->dev;
> >       struct fwnode_handle *fwnode = dev_fwnode(dev);
> >
> >       if (is_acpi_device_node(fwnode)) {
> >               dev_set_name(dev, "spi-%s", acpi_dev_name(to_acpi_device_node(fwnode)));
> >               return;
> >       }
> >
> >       if (is_software_node(fwnode)) {
> >               dev_set_name(dev, "spi-%s", fwnode_get_name(fwnode));
> >               return;
> >       }
> >
> > i.o.w. we don't need to have two ways of checking fwnode type and you may get
> > rid of unneeded variable, and always use fwnode name for swnode.
> >
> > ...
> >
> > > +   proxy->use_fwnode_name = chip->use_fwnode_name;
> >
> > Unneeded variable. See above.
>
> Hmm... I guess I was viewing this feature more as something that
> users would opt into. So other firmware mechanisms could use it
> if required, and so most swnode based controllers would still get
> caught by the standard naming at the bottom of the function.

software nodes can be used as a trampoline for old board files (in
order to make the driver cleaner and prepared for DT conversion of the
board file in question) or for quirks. In either case when we use them
we really want to have their data to be propagated unconditionally
(just based on the type), the (per-subsystem) gate is a carefully
placed mine on the minefield.

> From my perspective it will do what I need either way, so happy
> to update it to always use this for software nodes if consensus
> goes that way.

Not a maintainer here, ask Mark.

-- 
With Best Regards,
Andy Shevchenko